The significance of the stratigraphic record in Kents Cavern, Devon, United Kingdom, to the interpretation of the British Quaternary is confirmed on the basis of a thorough reexamination of the deposits in concert with 2 new Al-Be cosmogenic and 34 new thermal ionization mass spectrometry U-Th dates. The deposits show evidence of complex reworking in response to periglaciation, and the main flowstone deposit is a multilayered complex spanning marine isotope stage (MIS) 11–3. The lowermost unit of fluvial sands is Cromerian or older. The second deposit, a muddy breccia of surficial periglacial solifluction material containing Acheulian artifacts, entered the cave during MIS 12 from high-level openings to the west. Cave bears denned in the cave during MIS 11, the Hoxnian interglacial; their bones are capped by an MIS 11 calcite flowstone layer. From MIS 11 onward, each interglacial period and the warmer interstadial periods (MIS 11, 10b, 9, 7, 6b, 5, and 3) produced calcite flowstone deposition in the cave; MIS 9 was particularly active. Each glacial or stadial period (MIS 10c, 10a, 8, 6c, 6a, 4, and 2) caused periglacial activity in the cave, during which the thinner layers of calcite were fractured by frost heave and redistributed by solifluction. This sequence was interrupted during MIS 3–2 with the introduction of sandy and stony clastic sediments from entrances to the east, and finally cemented by the uppermost layer of MIS 1 flowstone. This is the first publication of well-dated and clearly documented evidence of frost heaving in interior cave passages. The Kents Cavern record of continuous, repeated sedimentation events followed by frost shattering and remobilization events over the past 500 k.y. is probably unique in the karst literature and establishes Kents Cavern as a site of international scientific interest.

Kents Cavern, Torquay, Devon, UK (NGR SX 934 642, Fig. 1), has been a focus of scientific interest for at least the past two centuries. The extraordinarily rich record of Pleistocene mammals and human artifacts has focused interest on the paleontology and archaeology of the site (e.g., MacEnery, 1859; Pengelly, 1869, 1884; Campbell and Sampson, 1971; Cook and Jacobi, 1998), while the sedimentological, climatological, and geomorphological information in the deposits has largely been ignored. Here we focus on the reconstruction of the Pleistocene depositional events of the past 500 k.y. that are recorded in this inter nationally important site.

The south Devon karst is one of the few limestone areas in Britain that was not directly affected by the Pleistocene ice sheets. Deposition of both clastic and crystalline material in the cave was abundant, such that much of the site was barely accessible to early explorers (MacEnery, 1859). The first systematic excavations began in the early 1800s; since then, the deposits have mostly been removed from the cave so that the cavity now exposed is considerably larger than that accessible in the early nineteenth century. The major excavation was carried out in 1868–1880 by W. Pengelly (see reconstruction of Pengelly's excavations in McFarlane and Lundberg, 2005). MacEnery (1859) and Pengelly (1884) described the basic sedimentological sequence, from oldest to youngest, as Breccia, Crystalline Stalagmite, Loamy Cave Earth, Stony Cave Earth, Granular Stalagmite, and Black Mold, and these terms have been retained by later workers. While this stratigraphy appears to be adequate for the 14C dated material, the clastic deposits, and the upper parts of the calcite (Fig. 2; Oxford Radiocarbon Accelerator Unit, 2006), it is insufficient to describe the Crystalline Stalagmite and hides a wealth of complexity. Proctor (1994) expended considerable effort to date the calcites by uranium series disequilibrium alpha counting (alpha U-Th) and by electron spin resonance (ESR), and continued this simple designation, with some minor variations. Proctor et al. (2005) added three thermal ionization mass spectrometric U-Th (TIMS U-Th) dates to the data set in an attempt to clarify the interpretation of the lowermost unit, the Breccia.

Improvements in the technology of both U-Th dating and the development of cosmogenic isotope dating for cave sediments now allow a more definitive dating of the older calcite and clastic deposits. The aim of this study is twofold: first, we clarify the depositional sequence and its paleoenvironmental context based on 34 new, high-precision TIMS U-Th dates; second, we present the first direct radiometric dates on the Breccia and its antecedent stratum, providing a well-supported time window for the middle Pleistocene fauna and human artifacts from the cave.

Geological Background

The geological setting was described by Durrance and Laming (1982) and Scrivener (1987). The cave is developed in the Middle–Upper Devonian Torquay Limestone (Fig. 1B). The underlying Nordon Slate is a gray, well-cleaved, calcareous mudstone with some bands of slaty limestone. This gives way to a complex of volcanic tuff, agglomerate, and lava interspersed with limestone, much of which is reefal in structure; the whole unit is designated as the Torquay Limestone. The overlying Upper Devonian Gurrington Slate is a gray-green and purple mudstone. These rocks were severely fractured and folded by the Variscan orogeny ca. 300 Ma. New Red Sandstones of Permian age unconformably overlie the uplifted and eroded post-orogenic surface. The Devonian rocks generally show substantial weathering underneath the sandstone beds.

Many relatively small caves have formed in the Torquay Limestone; Kents Cavern is the third largest, with <1 km of mapped passages. Kents Cavern preserves evidence of both phreatic and vadose activity, but phreatic features generally dominate. The cave probably formed in the early Pleistocene, but speleo-genesis may have involved some rejuvenation of late Carboniferous–Permian paleokarstic features (Campbell et al., 1998b). The geomorphological and hydrologic setting must have been different during speleogenesis, because the modern catchment for Kents Cavern, set high on a small hillside, is too small to provide substantial volumes of water.

The Quaternary deposits in the region include raised beach deposits, river terrace gravels, and periglacial deposits (Campbell et al., 1998a). The ice sheets of the Quaternary glaciations did not reach the Torquay district, but periglacial conditions were widespread during glacial periods (Cullingford, 1982). The two glacial ice sheets that came closest to Kents Cavern were the last, during marine isotope stage (MIS) 2, the Devensian, peaking ca. 20 ka, and MIS 12, the Anglian, ca. 430 ka (Fig. 1A). Periglaciation triggered extensive frost action and produced thick periglacial slope deposits or solifluction deposits (locally called “head”). The slates and mudstones were particularly susceptible to frost action and resulted in thicker than average head deposits (some sections documented as >20 m thick; Cullingford, 1982).

Sampling (Fig. 3) was undertaken with the support and cooperation of the cave owner and under permit from English Nature. Particular care and attention was paid to environmental impact, and the work was completed with minimal or no visible trace. Samples were taken of calcite that had already been damaged by historic blasting and excavations. In order to minimize the collection of undatable specimens, all potential sites were examined in the field for evidence of open-system behavior such as vugs, hiatuses, loose or friable or random fabric, recrystallization, or dissolution. Sample sites were recorded on site using photographs and photographic mosaics. In view of the complicated stratigraphy and in order to avoid subsequent misinterpretation of photographs, sequences were logged and photographs annotated in the cave using a portable computer.

26Al-10Be Dating

Recent advances in dating cave sediments have achieved considerable success with 26Al-10Be dating of buried quartz (Granger et al., 2001; Anthony and Granger, 2004). Cosmic ray secondary neutron bombardment of exposed or superficially buried quartz generates 26Al from silicon atoms and 10Be from oxygen atoms in a constant ratio of ∼6:1, regardless of absolute dose. After transport and burial in a cave, these nuclides decay at differing rates (half-lives: 26Al = 1.02 Ma; 10Be = 1.93 Ma), causing a shift in the nuclide ratio that can be used to date sediment burial ages in the range of ca. 100 ka to 5 Ma (Granger and Muzikar, 2001). Two materials were chosen as potentially amenable to this technique 01: the quartz sand from the basal Red Sands of the Gallery (cf. Campbell and Sampson, 1971) and quartz pebbles from the exposed Breccia in the Bear's Den and Labyrinth areas. Samples were prepared and analyzed at the Purdue Rare Isotope laboratory, Purdue University.

U-Th Dating

The calcite samples were dated by standard U-Th disequilibrium techniques (e.g., Ivano vich and Harmon, 1992). Specimens were sliced into 2-mm-thick slivers and examined under a binocular microscope with back lighting. Only the cleanest parts were used for dating. All visible traces of detritus, vugs, or intercrystallite voids were removed with a dentist's drill under the microscope. The very low U content (0.05 ± 0.02 ppm) dictated the use of relatively large sample sizes (∼2 g). Samples (n = 45) were ultrasonically cleaned, ignited for 5 h at 875 °C to remove organics, dissolved in HNO3 and spiked with 233U-236U-229Th tracer. Apart from the samples chosen for isochron dating (see following), no sample showed visible detrital contamination. U and Th were coprecipitated with iron hydroxide, and purified twice on anion exchange columns (Dowex AG1-X 200–400 mesh).

Measurement of U and Th isotopic ratios was mainly done with TIMS, 14 using the VG 354 TIMS at McMaster University, Hamilton, Ontario, and 19 using the Triton TIMS at the Isotope Geochemistry and Geochronology Research Centre, Carleton University, Ottawa, Ontario. The 8 samples for isochron dating and 4 repeats were measured using the multi-collector inductively coupled mass spectrometer (MC-ICP-MS) at Géotop, University of Quebec at Montreal, Quebec. Each suite of measurements was accompanied by the processing of uraninite in secular equilibrium to ensure accurate spike calibration and fractionation correction. Precision of isotopic ratio measurement is limited by extremely low U content and by high 232Th content. The typical 2σ error of the TIMS measurement of 234U/238U is 0.12% and of 230Th/234U is 0.46% (TIMS instrumental reproducibility on spiked uraninite standards is 0.06% for 234U/238U and 0.11% for 230Th/234U). The precision on the resultant ages varies with age; the average 2σ precision on these mid-Pleistocene dates is 2.2% 02.

Three of the hand specimens were sampled twice. CDF and B20 were measured on the McMaster TIMS and CDF2 and B20R on the Géotop MC-ICP-MS. The coincidence of the numbers (within 2σ) suggests that isotopic measurements on TIMS and ICP are com patible. Samples CEB and CEBR were not from the same section of hand specimen: CEBR was deliberately chosen from a different part in the hope that it would not be leached. Similarly, LOBO and LOAN were deliberately chosen from different parts of the hand specimen because they had visible detrital contamination.

Two of the samples with highest detrital thorium concentration, SW1-O and LB-O, were sampled 4–6 times for isochron dating (Schwarcz and Latham, 1989; Ivanovich et al., 1992); each subsample was chosen for its visibly differing detrital content. The principle of isochron dating is that the subsamples are deposited at the same time but with differing amounts of detrital contamination. The carbonate fraction 230Th is presumed to be constant for all sub-samples, but the detrital fraction 230Th varies; 232Th is used as an index of detrital content. The slope of the regression line of 234U/232Th against 238U/232Th gives the 234U/238U, and the slope of 230Th/232Th against 234U/232Th gives 230Th/234U at zero detrital contamination (Fig. 4). These ratios are then used to calculate the isochron age.

A 230Th/232Th ratio of <20 is normally considered to indicate detritally contaminated material (e.g., Schwarcz and Blackwell, 1992). However, isochron dating is prohibitively expensive for many samples. A simplistic way to adjust the 230Th/234U to account for detrital contamination is to estimate an initial 230Th/232Th activity ratio for the detrital fraction. For example, Kaufman and Broecker (1965) used a value of 1.7 from measurement of modern detritus. In our case the sample SWO had an age of 183 ka, but a 230Th /232Th ratio of 3. It was then dated by isochron on another four fractions (SW1-O) to 152 – 30/+45 ka. If this is taken as the more correct age, then the 230Th/234U ratio for SWO can be adjusted to give an age of 152 ka using an initial 230Th/232Th ratio of 1.00. This estimate for initial 230Th/232Th was then used for all the samples with a 230Th/232Th ratio of <20. It is clearly only an approximation, but is an improvement over the alternate practice of simply using a standard value. If the estimated initial 230Th/232Th ratio is low, and/or if the measured 230Th/232Th ratio is high, the adjustment does not alter the original date.

In 02, the age for the samples with detrital contamination is shown first as the simple calculated age; beside it in parentheses is the adjusted age; the age used is highlighted in bold. In most cases, the adjusted age is within error of the original age, so there is no justification for using anything other than the original age. Only three of the dates we have used are the adjusted ones.

Initial 234U/238U ratios for the in situ flow-stone samples are all within a narrow range that gradually decreases over the course of the Pleistocene, a very common pattern explained by depletion of the more soluble 234U in the overburden (Fig. 5). The values for the materials that are not in situ, but rather have been transported from another part of the cave, are well above the 3σ range. The one exception is sample HLC2, flowstone that is in situ, but has an initial 234U/238U ratio below the 3σ range. This sample is from a high part of the cave, close to the surface with only a thin overburden.

Al-Be Dating

An analysis of quartz sand from the basal Red Sands of the Gallery (Figs. 6A, 6B) and quartz pebbles from the exposed Breccia in the Bear's Den–Labyrinth area (Fig. 6C) yielded finite 26Al-10Be dates, albeit with very broad error margins 01. Unfortunately the Red Sands are only exposed in this one site and the relationship with Breccia is not explicit. The Gallery Red Sands yielded a date of 2.32 ± 1.74 Ma (1σ range: 0.58–4.06 Ma) and the Breccia yielded a date of 0.95 ± 0.55 Ma (1σ range: 0.4–1.5 Ma).

U-Th dating

Data are presented in 02. We present the results from each site. The cave stratigraphy is complex, such that no one site offers a continuous sequence of events; instead the sequence must be reconstructed from the partial information from many sites.

Hedges Boss

Of all the sites, Hedges Boss (Fig. 7A) has the simplest stratigraphic relationships. The Breccia (>2 m thick here) is capped by ∼10 cm of orange-pink, laminated calcite (sample HB-2B) that was dated to 408 +15/–14 ka, early MIS 11. The sequence continues with a thin layer of red, detritus-rich calcite that in this cave indicates a hiatus in deposition and/or very slow deposition. This is capped by ∼10 cm of white, laminated calcite with several poorly expressed hiatuses, and a thin layer that is currently active (MIS 1). Figure 7B shows the diagrammatic interpretation of this section. We have no other dates from this site; therefore the layers from MIS 9, 7, and 5 are presumed, based on data from other parts of the cave. Proctor et al.'s (2005) sample KC–90–2, taken from just below the hiatus (a position confirmed by the large saw-cut remaining from their sampling), shows evidence of recrystallization. Our sample was not taken close to the hiatus and shows the clearly defined fine growth laminations of the original fabric. Thus there is no reason to suspect this date.

Bear's Den

The stratigraphic relationship of the Breccia and the calcite is not simple here (Fig. 8A). Sample BD-1 (326 ± 7 ka, MIS 9) is a clean, orange-pink, laminated calcite from immediately above the Breccia-calcite contact, ∼6 m to the right of the Bear's Den Boss (inset Fig. 8A, placed in correct stratigraphic position). Sample BD-2 (439 +23/–19 ka, MIS 11) is from a layer of white calcite coating bedrock that curves out into the Breccia. Proctor (1994) interpreted this as a calcite vein intercalating between two beds of breccia. Having cleaned the face with wire brush and water, we interpret this calcite as simply filling a narrow gap between the Breccia and the rock, probably created by dripping, and thus postdating the Breccia (Fig. 8B).

Water Gallery

Because so little material remained, only one sample (WG1U: 307 ± 5 ka, MIS 9) was taken of the calcite layer immediately above the Breccia. All that is left of the white, laminated calcite flowstone is a flake against the rock but separated from rock by a thin film of mud, which we interpret as Breccia remains.

Clinnick's Gallery

Here the stratigraphic sequence between the two sampling locations is not continuous. Figure 9B (shown in correct position relative to the main sampling site, Fig. 9A) shows the contact with Breccia. Immediately on top of the Breccia is ∼25 cm of soft, vuggy calcite. We sampled the solid, cream colored, opaque calcite above this (CG-TB:193 ± 1 ka, MIS 7). The calcite is separated from the rock face by a thin film of mud, which we interpret as the remains of Breccia. No hiatuses are apparent between the contact and the sampling site; we thus assume that the entire calcite layer was deposited during MIS 7.

The focus for the main site was on the broken flowstone in the middle (Figs. 9A, 9D). The sequence shows that the upper layers of white, laminated calcite (sample G2-BW, 79 ± 3 ka, MIS 5a) were fractured and displaced, and cemented in place by thin layers of red calcite (sample G2-TR, 47 ± 3 ka, MIS 3). These were in turn fractured and cemented in place by the topmost layer of white vuggy Granular Stalagmite of MIS 1 age. Figure 9C shows the diagrammatic interpretation.

High Level Chamber

In the High Level Chamber (Figs. 10A, 10B), the remains of calcite flowstone with a natural fracture surface (rather than the typical clean and scraped surface from excavation) are plastered to one side of the passage at ∼1.5 m above the present floor. The lowermost 10 cm of flowstone is not amenable to dating (labeled as sugary calcite in Fig. 10A). Sample HLC-2 (121 ± 1 ka, MIS 5e) comes from the first layer of solid, laminated calcite. No hiatuses are apparent between the contact and the sampling site; we thus assume that the entire calcite layer was deposited during MIS 5e. On close examination, it is apparent that it had been fractured and cemented with overgrowth of red, muddy calcite. The site is further complicated by the remains of white, vuggy calcite both above the 5e deposit and just below it. Proctor's (1994) alpha-counted U-Th date on this material (the lower layer) is 53 +6/–4 ka, placing it in MIS 3.

In-Between Boss

This boss, between Inscribed Boss to the north and Hedges Boss to the south, unnamed on the Proctor and Smart (1989) survey, is shown in Figure 11A (a circumferential mosaic). Much of the most clearly exposed face to the left could not be sampled for aesthetic reasons. Of the two samples taken in direct contact with the Breccia, CEB was leached and CI-HB dated to 289 +20/–17 ka, MIS 9. All three dates from this lowermost block of clean, white calcite are not statistically separable, showing that deposition was rapid in MIS 9 (CI-G, 312 +10/–9 ka and CI-F, 310 ± 6 ka). The thinner parts of this layer to either side of the main boss are cracked (see following 4). It was originally presumed that sample CI-E2 (220 +4/–3 ka, MIS 7) would be stratigraphically equivalent to CDF (150 ± 5 ka, MIS 6b), but the dates prove otherwise. The remaining samples show a clear progression through MIS 5 (CCD, at 100 ± 3 ka, CCT, at 95 ± 4 ka, CBN, at 93 ± 2 ka, and CAT, at 80 ± 1 ka). The overlying drapery is currently active and assigned to MIS 1. Figure 11B shows our reconstruction of the sequence of deposits.

Labyrinth

This thick layer of flowstone emerging from the Little Oven Exit that originally spilled into the Labyrinth (Figs. 12A, 12B) was difficult to sample. Much of the calcite is vuggy, and appears to have been deposited in shallow standing water. The dates (LB-3: 296 +9/–8 ka; LB-5: 293 ± 6 ka; LB-4: 315 +7/–6 ka) suggest rapid deposition in MIS 9. Two samples were taken from within the Breccia. The first, a broken slab of flowstone, LB-O, was isochron dated to 311 +28/–22 ka, MIS 9. The second, a piece of stalagmite, broken and embedded in the Breccia, sampled ∼5 m to the left of this site (sample B20, shown in correct relative stratigraphic position in Fig. 12B), was dated as 210 ± 2, MIS 7. These last two have initial 234U/238U ratios well outside the 3σ range for the in situ flowstone (Fig. 5, open circles), suggesting that they originate from elsewhere in the cave.

Southwest Chamber

We sampled from both the northwest (Figs. 13A, 13B) and southeast (Figs. 13C, 13D) sides of this passage. The northwest side shows a simple sequence: SW-3 (293 +14/–12 ka, MIS 9) in direct contact with Breccia, up through SW-5 (245 +3/–2 ka, MIS 7) and SW-6 (238 ± 4 ka, MIS 7), and red vuggy calcite on top that was not amenable to dating. On the southeast side, the calcite remains plastered high on the wall consist of the lowermost layer of rapidly deposited, dendritic fabric, full of intercrystallite voids and not amenable to dating; an intermediate layer of white, laminated calcite (WUXU, dated imprecisely to 417 +104/–52 ka, MIS 11); and white laminated calcite (SWAN, 306 ± 4, MIS 9). The lower part of the southeast side shows a sequence of calcite in contact with breccia (SW1M, 352 ± 6 ka, MIS 10b); a hiatus; a thin layer of redder calcite (WTXT, 132 ± 6 ka, MIS 5e); and vuggy, Granular Stalagmite (MIS 1). The sample taken from within the Breccia, of a fractured flowstone slab, SW1-O, was isochron dated as 152 +45/–30 ka, MIS 6b, and shows an initial 234U/238U ratio outside of the 3σ range (Fig. 5), indicating that it probably originated in another part of the cave.

In order to understand the evidence of Quaternary events in Kents Cavern it is important to be aware of some of the complexities of cave sedimentology in general. The age of any one deposit cannot usually be predicted from its geomorphological position because the principle of superposition does not necessarily apply in caves. In some sections deposition appears to have obvious hiatuses, but dating shows that they are in reality simply shifts of the drip point (e.g., some parts of In-Between Boss). Reworking of nonindurated sediments may complicate temporal relationships: we believe that the clastic sediments in this cave have been repeatedly remobilized since initial deposition, confusing the sequential relationships.

It is also important to be aware of the problems of inconsistent application in the literature of British stage names and associations with the marine oxygen isotope record. The mid-Pleistocene record from British sites is somewhat hazy because so many of the sites cannot be radiometrically dated and the assigned age depends on a combination of paleontological, archaeological, and palynological remains, on stratigraphic position, and on sedimentology.

Differentiation of the post-Anglianinter glacial sites (e.g., Purfleet and Hoxnian Interglacials) normally relies on aminostratigraphy. However, McCarroll (2002) suggested that this method is not sensitive enough to allow confident separation of populations into different interglacials. It is now widely accepted that sites attributed to the Hoxnian on the basis of pollen spectra may represent more than one warm stage (Scourse et al., 1999; Dowling and Coxon, 2001; Thomas, 2001). Schreve and Thomas (2001) suggested that two episodes are recorded, each with a Hoxnian-type pollen signature. As new data are published, the controversy lessens. For example, Grün and Schwarcz (2000), from U-series/ESR ages of 403 +33/–42 ka on teeth, placed deposits of the type locality of the Hoxnian Interglacial in MIS 11. Rowe et al. (1999), based on careful U-series dating of lake sediments, also assigned the Hoxnian to MIS 11.

The Anglian is the most widely recognized event in the mid-Pleistocene of Britain, yet its age is not conclusively established. The MIS 11 date for Hoxnian deposits from Rowe et al. (1999) also assigns the Anglian glacial deposits that underlie the Hoxnian deposits, with no evidence of any significant break in deposition, to MIS 12. The majority of publications refer the Anglian to MIS 12 (Schreve and Thomas, 2001).

Cromerian sites from Britain have, in the past, been assumed to represent a single interglacial stage, whereas in the Netherlands the Cromerian Complex has incorporated four interglacials and their intervening cold stages (Preece, 2001). Recent molluscan evidence has supported several distinct stages (Preece, 2001; Schreve and Thomas, 2001). An age of MIS 13 is generally agreed as the younger limit, but there is no consensus about the age of the start of the Cromerian; Parfitt et al. (2005) proposed at least MIS 17 for the Cromer Forest bed, and by implication, the Cromerian type site at West Runton.

In the following discussion we assume that the Hoxnian stage correlates with MIS 11, the Anglian stage with MIS 12, and the Cromerian is MIS 13 to at least MIS 15.

The distribution of the standard four units, i.e., the Breccia, the Crystalline Stalagmite, the Cave Earth, and the Granular Stalagmite, was described by Pengelly (1884), Keith et al. (1931), Campbell and Sampson (1971), Proctor (1994), Straw (1997), and Proctor et al. (2005), but without a reliable temporal framework for materials outside the range of 14C dating. Proctor (1994) provided detailed descriptions of facies. Here we focus on the information from the new dates and present a new interpretation of the stratigraphy.

Age and Emplacement of the Red Sands

A sometimes-overlooked deposit, the Red Sands, has been variously interpreted. Keith et al. (1931) reported the Red Sands as the basal unit below the Breccia in the Water Gallery, but this exposure is now buried by a cement pathway and cannot be reexamined. Keith et al. (1931) and Campbell and Sampson (1971) correlated these sands with the Red Sands currently exposed in the Gallery, but Proctor (1994) interpreted the unit as a wash facies of the loamy Cave Earth (well dated by 14C as late MIS 3, Fig. 2). The Gallery exposure remains accessible, in a pit dug into the floor of the Gallery, but the relationship with the Breccia is ambiguous because the sands are separated from the overlying Breccia by a substantial vacuity. The Red Sands (Fig. 6A) comprise a basal unit of horizontally bedded mud; a middle unit of bedded sands that dips steeply ∼30° away from South Entrance; and an upper unit of poorly bedded gravel, the clasts of which are subrounded, poorly sorted, mostly 1–2 cm but as much as 5 cm in diameter. We interpret this as a simple fluvial sequence of low-water mud, current-bedded sands, and high-water river gravel (Fig. 6B).

We address the relationship of the Red Sands to the Cave Earth and the relationship of the Red Sands to the Breccia. Even with the very large 1σ error, the probability that the 26Al -10Be burial age of the Red Sands is younger than the earliest possible age on the Cave Earth (i.e., end of MIS 5a, 79 ka) is only 0.09 (Monte Carlo simulation, 10,000 iterations). Thus Proctor's (1994) hypothesis that the Gallery sands are a remnant of the Cave Earth can be rejected.

The 26Al-10Be data support a significantly greater age for the Gallery sands versus the Breccia, and support Campbell and Sampson's (1971) view that the Gallery sands are basal to the Breccia and synonymous with Keith et al.'s (1931) Red Sands, making them the oldest dated deposits preserved in the cave. In addition, examination of the 26Al nuclide concentrations in Gallery quartz versus Breccia quartz reveals that they differ by more than 4 standard deviations 01, demonstrating that the exposure history of these materials has been quite different.

We suggest that the Red Sands are simple fluvial deposits indicating a flow with a high clastic load, of variable discharge, moving up South West Chamber and diverging through the Gallery toward the Long Arcade and North Entrance. While no definitive date can be assigned (in view of the high error margin of the cosmogenic date), it would be reasonable to assume that the Red Sands are Cromerian (MIS 13–15, a long period of predominately temperate conditions), and represent the final stages of vadose activity after the phreatic activity that carved the primary cave passages.

Age of the Breccia

The 26Al-10Be burial date on the quartz grains within the Breccia gives an age estimate of mid-Pleistocene. The oldest of the U-Th dates on the capping flowstone offers an upper window. However, the possibility of a potential hiatus between sediment deposition and calcite deposition must be acknowledged in view of Stock et al.'s (2005) findings that U-Th dates on speleothem overlying clastic sediments are often considerably younger than cosmogenic 26Al-10Be burial dates on the sediment. In Kents Cavern, the Breccia contains additional evidence of the time of its formation in the form of sedimentological characteristics and paleontological remains.

We dated 11 samples of calcite flowstone in direct contact with the Breccia: 2 of these yielded MIS 11 dates, 5 had MIS 9 dates, 2 had MIS 7 dates, and 1 had a date of MIS 5. The simplest interpretation from the capping calcite is that the Breccia formed before MIS 11 and the younger dates represent hiatuses of various lengths. Proctor et al. (2005) tried to date the Breccia using the calcite cap, but they took the average (MIS 9) as an indication of minimum age. Any averaged time series will yield a mean that is significantly younger than the actual age of the initiation of the series.

Sedimentological Evidence

Sedimentologically, the Breccia is a poorly sorted diamict of angular to subangular clasts (red sandstone, siltstone, slate, quartz, and rarely limestone) in a matrix of red mud (Fig. 6C). Proctor (1994) noted that the clasts are as large as 20 cm and are matrix supported, and that little or no fabric is visible in the generally homogeneous deposit. This material is typical of frost-shattered regolith or head deposits that abound on the hillsides of the area as a result of former periglacial activity (Cullingford, 1982; Scrivener, 1987; Croot and Griffiths, 2001). This requires a long period of cold conditions. The Cromerian represents a long period of temperate conditions from MIS 13 to at least MIS 15 (474–620 ka; Bassinot et al., 1994), but also incorporates the full glacial at MIS 16 together with at least one preceding interglacial (Parfitt et al., 2005). The Breccia is unlikely to date to the early Cromerian: if the cave were open and the Breccia had been produced in the MIS 16 glacial, the Breccia and the cave would have to have remained completely unaffected by the 146 k.y. of Cromerian temperate conditions and another 46 k.y. of Anglian glacial conditions until the initiation of calcite deposition in MIS 11. Thus we conclude that the Breccia must have been produced in a post–MIS 16 cold stage. The only stage that is cold enough between the oldest date on the calcite, MIS 11, and MIS 16 is MIS 12, the Anglian glacial period (MIS 14 was too warm). The question is not resolved by other head deposits, because so few have been reliably dated. Bates et al. (2003) indicated that most are from MIS 4 and MIS 2, some are from MIS 6 or MIS 8, but in a few places with good dating control, head deposits dating to the marine oxygen isotope stage 12 can be recognized. For example, head deposits from cold stages back to MIS 12 have been found (with ages ranging from MIS 11, on amino acid evidence, to MIS 13, on mammalian biostratigraphy) in the Hampshire-Sussex coastal plain. Murton and Lautridou (2003) also dated periglacial deposits along the English Channel coastlands (by radio-carbon, luminescence, and mammalian biostratigraphy); they place most in MIS 2, some in MIS 6, and some in other cold stages.

Paleontological Evidence

Although their U-series ages provide only an upper window of MIS 9, suggesting that the Breccia could represent either MIS 12 or MIS 10 deposition, Proctor et al. (2005) argued that the faunal remains in the Breccia suggest a late Cromerian age, consistent with many examples of well-established pre-Anglian faunas in UK sites and in the Netherlands. The paleontological evidence comes from both the rodent remains (generally disseminated throughout the Breccia) and bear remains (generally at the top of the Breccia).

Evidence From the Rodents

Pengelly's notes (1868–1880) do not address the microvertebrate fauna of Kents Cavern, but workers have identified sparse remains of the voles Pitymys gregaloides, Arvicolagreeni” (= A. cantiana; Sutcliffe and Kowalski, 1976) (Campbell and Sampson, 1971), and Microtus oeconomus (Proctor, 1994). The evolution of voles is of great importance in the middle Pleistocene biostratigraphy of Europe, and provides time constraints on the Kents Cavern Breccia deposits. The British extinct water vole lineage consists of the chronospecies Mimomys pliocaenicus (early Pleistocene), M. savini (late-early through early-middle Pleistocene), and A. cantiana (late-middle Pleistocene to early-late Pleistocene) (Lister et al., 1990). Neither Mimomys species is known from the Kents Cavern Breccia. The appearance of unrooted molar teeth characteristic of the genus Arvicola is dated to MIS 11 in continental Europe (Pevzner et al., 2001), but to the late Cromerian (Sutcliffe and Kowalski, 1976) or immediately pre-Anglian (Andrews, 1990) at the Westbury site in Britain. This sets the maximum age limit for the Kents Cavern Breccia fauna at MIS 13. The extinct Pine vole, Pitymys gregaloides, is also present at Westbury but has not been found in any British Hoxnian (MIS 11, sensu Schreve, 2001) site (Sutcliffe and Kowalski, 1976), and is generally presumed to be indicative of pre–MIS 11 age.

Thus the rodent fauna are MIS 13 in age (we agree with Proctor et al., 2005). The deposit in which they are incorporated may also be MIS 13, but could equally be younger. If we accept that the Breccia represents a cold-climate (periglacial) deposit, and that the oldest calcite capping is MIS 11, then the only possible age of emplacement would be MIS 12.

Evidence From the Bears

The Kents Cavern Breccia is most notable for yielding large numbers of teeth and bones attributed to the cave bear, a term that is applied to animals in the Ursus saviniUrsus deningeriUrsus speleaus chronospecific lineage. The exact position of the Breccia bears along this lineage has been ambiguous, in part because the material has not been formally reviewed and also because the systematics of European middle Pleistocene bears has not been fully resolved. Bishop (1982) placed the Westbury (MIS 13) bears at full U. deningeri grade. Schreve (2001) considered bears of full U. speleaus grade to be Hoxnian (MIS 11) in age. A preliminary analysis of the Breccia bear teeth (McFarlane et al., 2006) indicates that they are of an advanced U. deningeri–early U. speleaus grade, consistent with an age of MIS 12–11, but clearly younger than MIS 15 (U. savini; Kurtén, 1968), and probably younger than the Westbury MIS 13 bears. A small proportion of generally less well preserved bear material is found disseminated throughout the deposit, but the majority of the bear material is found within the uppermost 25 cm, immediately below the calcite cap, indicative of a superficial emplacement, so much so that MacEnery (1859, p. 27; our brackets) reported “The first flag [of overlying flowstone] that was turned over exhibited, in relief, groups of skulls and bones adhering to the stalagmite.” The majority of the bones are from bears using Kents Cavern as a hibernaculum, on top of the Breccia, and the bones are incorporated into the topmost muddy layer. Thus most of the bear material postdates the MIS 12 emplacement of the Breccia.

Evidence From the Artifacts

Much of the historic interest in Kents Cavern has resulted from the recovery of very early Acheulian flint and chert artifacts. Acheulian biface tools are known elsewhere in Britain from Boxgrove (MIS 13) and Pakefield (MIS 17; Parfitt et al., 2005). Campbell and Sampson (1971) noted that Pengelly recovered these artifacts throughout the thickness of the Breccia, including its lowest levels. Moreover, the largest number of artifacts (31%) came from the “4 ft” level within the Breccia. Several authors (Campbell and Sampson, 1971; Cook and Jacobi, 1998) have commented on the poor condition of the artifacts, which show evidence of both rolling and rotting, consistent with long exposure on the surface and subsequent entrainment in the Breccia debris flow.

The archaeological and paleontological evidence suggests that the Acheulian artifacts accumulated on the surface in pre-Breccia time, and were subsequently carried into the cave entrained in the Breccia debris flows. In contrast, the bear remains are both spatially and taphonomically consistent with having been derived from a bear hibernaculum in the cave, on top of and postdating the Breccia. Subsequent remobilization of the Breccia (see following) has incorporated some of this bone material in its upper layers. The provenance of the rodent remains is not known with certainty, and these animals may include specimens both coeval with and postdating the emplacement of the Breccia.

Emplacement of the Breccia

Generally the Breccia shows very little internal structure. It is clearly not fluvially emplaced, but rather is a mass-movement deposit. Proctor (1994) and Straw (1997) suggested that it is a debris flow, on the basis of the thick, structureless beds with poor sorting, chaotic to sub horizontal clast orientation, and matrix support. Collcutt (1986) defined debris flows as water-saturated materials moving at speeds detectable to the observer, up to tens of kilometers per hour. He considered them to be common events in caves and the dominant mechanism for lateral displacement of deposits. Initiation of movement requires slopes of at least 20° and a sudden input of large quantities of water. Thus, Proctor (1994) suggested that the Kents Cavern Breccia debris flows would have required a high water content and that emplacement in the cave was accompanied by small streams (a view reiterated in Proctor et al., 2005). This suggests a warm or at least warming climate. Collcutt (1986) noted that some sorting occurs during debris flow, the larger particles moving to the sides and base; that for high-water-content flows there is a tendency for orientation of particles with flow; and that larger clasts or artifacts and faunal remains may be significantly worn and damaged by the flow. While there is little clear evidence of sorting, damage to the artifacts disseminated throughout the Breccia (Cook and Jacobi, 1998) supports this view. However, it is apparent to us that the features of this deposit could equally be interpreted as a solifluction flow (see following). Croot and Griffiths (2001) observed that the head material on the surface has been moved downslope by solifluction, or by rapid “slushflows” and slides, or by torrential flash-flood events. We suggest that the head material could equally have moved into the cave by such processes.

Bertran et al. (1997) defined solifluction as slow mass movement caused by freezing and thawing, combining small-scale downslope frost creep and viscous flow from rapid release of meltwater during thawing. It produces lobes and sheets, usually with a preferred clast orientation parallel to slope. Debris flows are rapid mass movements of poorly sorted solids and water, usually triggered by rain events but also reported as resulting from rapid melting of ground ice triggering retrogressive thaw slumps. Fabrics for both these deposits have many analogies: both show clast orientation parallel to slope, and sometimes imbrication in clast-rich deposits. Data on clast orientation show that debris flows and solifluction deposits are not clearly differentiated, having a similar range of a-axis vector magnitudes. Both may exhibit simple shear deformation. Some differences between the two deposits can be seen, in that fabric strength is often lower in debris flows than in solifluction deposits, and that fabric shapes for debris flows are usually moderately developed girdles, whereas in solifluction deposits clusters and girdles occur in equal proportions. Millar (2006) observed that debris flows often produce imbricate fabric depending on the clast frequency and slope, but in solifluction the materials behave as a viscous fluid and may also produce imbricate fabric, depending on clast frequency and slope.

Proctor's (1994) description of breccia fabric is not detailed enough to allow application of Bertran et al.'s (1997) criteria. While the mode of emplacement of the Breccia does not really matter to its subsequent history, we suggest that the Breccia is a standard head deposit that was moved into the cave (through suffosion dolines, such as the Swallow Hole Gallery, and/or rifts) either by solifluction or by debris flow. Collcutt (1986) stated that excavators were too ready to blame processes such as cryoturbation. While this many be true, we suggest that not all matrix-supported, relatively homogeneous deposits in caves must necessarily now be reinterpreted as water-saturated debris flows, and that solifluction clearly continues to produce sediments with features similar to those in the Breccia.

The time frame available for breccia emplacement was relatively narrow: the head deposit that is the basis of the Breccia was formed by frost action during MIS 12; the bears moved in and used the top of the Breccia for a den; the top of the Breccia was partly encrusted in calcite during MIS 11 while the bears continued to use the area. Thus the bulk of the Breccia had to have been emplaced at the end of MIS 12 before the bears moved in. If the Breccia is a wet debris flow, then the time for emplacement had to have been during termination V or very early MIS 11.

Post-Breccia Calcite Flowstone

The appellation of all the Breccia-capping calcite as the Crystalline Stalagmite implies a simple sequence of synchronous events. In fact, except for the relatively clear case of Hedges Boss, the relationship of the Breccia and the overlying flowstone is rarely simple and rarely synchronous.

In the Bear's Den Boss, the visible stratigraphy suggests a lower layer of breccia, an overlying layer of flowstone, a second layer of breccia, and a cap of flowstone. Proctor et al. (2005) suggested that the lowest layer of flow-stone was broken up, accompanied by local reworking of the Breccia. An alternative view is that the lowest layer simply represents an exposed flowstone ledge (typical of the edge of a boss) that was buried by postdepositional movement of the Breccia creating the complex pattern of breccia-rock-ledge that is outlined in Figures 8A and 8B.

While the first breccia-topping calcite deposit was from MIS 11, the majority of dates indicate a significant episode of calcite deposition in MIS 9, e.g., almost 1 m of MIS 9 flow-stone in the In-Between Boss and in the Labyrinth. There was an even longer hiatus between breccia emplacement and calcite capping in other areas. The history of the breccia-calcite contact from Clinnick's Gallery (Fig. 9B) connects with that from Rocky Chamber, at its northern end. The absence of any calcite capping the Breccia until MIS 7, and the presence of red mud lining the junction of the calcite and the wall, suggests that the passage was completely blocked by breccia. The occurrence of previously unreported paragenetic pendants and anastomoses in the roof of Rocky Chamber (Fig. 14A), together with the remains of a fragment of breccia in the roof (Fig. 14B), suggests that Rocky Chamber was also filled to the roof with breccia. The paragenetic erosion suggests fluvial action after emplacement of the MIS 11 Breccia and before deposition of MIS 7 calcite. We suggest that the most likely time was the MIS 9 interglacial period. The delay in deposition of the calcite cap in High Level Chamber (Fig. 10A) until MIS 5 is also probably explained by the passage being largely blocked by breccia, although we observed no direct evidence of paragenesis here.

Southwest Chamber (Fig. 13) shows evidence of an even more complex suite of events. The high-level remnants of MIS 11 and MIS 9 flowstone indicate that the Breccia must have been at least 2 m deep in order to provide a substrate on which the flowstone grew. This must have been subsequently largely removed (in MIS 10) to allow deposition of the MIS 9 flowstone at the lower level. Drips from the roof continued to deposit calcite on top of the original MIS 11 layer (now a false floor or ledge), but also deposited calcite on top of the newly eroded Breccia at the same time. Thus in a single passage the thin remnants of flowstone represent two layers that were contemporary but vertically separated by several meters.

The evidence is clear that the Crystalline Stalagmite represents a complex of flowstone layers, and that each part of the cave requires detailed study. It is also relevant to note that the appellation Granular Stalagmite is not really adequate to represent the complex of Holocene calcite deposits. The cave opened up much more in the Holocene, so that deposition in the entrance zones and main passages was thick, porous, and vuggy, as is characteristic of rapid, evaporative deposition. However, Holocene deposits in less open parts of the cave are crystalline in fabric and still active.

Sequence of Events

Any scenario for the sequence of events must explain the formation and emplacement of all the sedimentary units, but it must also explain two unusual observations. The first is the evidence for widespread cracking of flowstone layers and the second is evidence for the incorporation of fractured material of younger ages within the upper layers of the Breccia.

Cracking of Flowstone

Many examples can be seen where a layer of flowstone is cracked and shifted, and the cracks are cemented with a thin coating of younger calcite. Most of the evidence of cracking was removed during excavations, but Straw (1997) reconstructed, from Pengelly's original reports, the distribution of Crystalline Stalagmite. Pengelly differentiated between Crystalline Stalagmite that was cracked in situ and Crystalline Stalagmite that was cracked and relocated, and was found as detached fragments in the Cave Earth. The in situ cracking was centered on the Bear's Den, Labyrinth, Cave of Inscriptions, and Clinnick's Gallery. The relocated fragments found in the Cave Earth were moved from the Vestibule southward and into North and South Sally Ports.

Only one attempt to date the time of cracking has been published. Proctor (1994) was able to delimit the date of cracking of the edge of Bear's Den Boss to after 115 ka (±4 ka), but this example has no younger overlying intact calcite that could provide an upper age limit. We studied several examples. The best is from Clinnick's Gallery (Figs. 9A, 9B, 9C). We dated the white flowstone layer that was cracked and the thin red calcite that coated the broken fragments. This gave a window for cracking and recementation of 79 ±3 ka (the end of MIS 5a interglacial) to 47 ±3 ka (the MIS 3 interstadial). Observant readers will note the high detrital content of the uppermost red calcite coating and the adjusted age of 20 ka. However, a further constraint on the time window for cracking is the subsequent deposition of Cave Earth that has been 14C dated as ca. 23–35 ka (Fig. 2). Thus the original age is the more likely, and the timing for cracking most likely to be MIS 4, the first cold stage of the Devensian glaciation. We also have to fit into this timing the cracking of the red calcite and its subsequent cementation with Granular Stalagmite, well dated by 14C to MIS 1, ca. 4–16 ka (Fig. 2). The second episode of cracking must have occurred between MIS 3 and MIS 1.

The evidence from High Level Chamber suggests that the MIS 5 flowstone was fractured and largely removed, and the Breccia surface level lowered, before the subsequent deposition of MIS 3 calcite under and slightly overlapping the MIS 5 flowstone. This cracking must also have occurred during MIS 4.

Another example of cracking that we dated is from the In-Between Boss, but the evidence is not so clear cut. Here the thick basal flowstone that was dated to MIS 9 was cracked and lined with red calcite, but only in its thinnest part, where it is ∼20 cm thick. At the left side of the boss the basal layer of MIS 9 calcite is cracked and overlain by MIS 7 calcite (sample CI-E2 220 +4/–3 ka). The cracking here must have occurred during MIS 8. However, at the right side of the boss, the layer above the red calcite that cements the cracks in MIS 9 material is dated as MIS 6b (sample CDF, dated twice, 150 ± 5 ka and 148 ± 9 ka), giving a wide window for cracking from MIS 9 to MIS 6b. While the evidence is not unequivocal, we argue that if the cracking on this side had occurred during MIS 8, then we would expect the crack to have filled with clastic material or calcite. The fact that the filling material dates to MIS 6b suggests that the cracking on this side had occurred immediately before this, perhaps during MIS 6c, the first cold period of MIS 6.

The evidence is for several episodes of cracking, and each of the times indicated, MIS 8, MIS 6c, MIS 4, and MIS 2, suggests a cold period. Further evidence for cracking is discussed in the following.

Incorporation of Younger Material

The second of the unusual events that must be explained in any sequence of events is the incorporation of fractured material of younger ages within the upper layers of the Breccia (in addition to blocks of angular bedrock). The three we sampled yielded dates of MIS 7, MIS 6b, and MIS 9 (B20—210 ka, SWO—152 ka, and LBO—311 ka). The standard explanation for mixing of materials of different ages is reworking of the sediment. Thus the Breccia (at least the surface few decimeters) must have been remobilized some time after its original emplacement.

The evidence for postdepositional movement of the Breccia is clear. MacEnery (1859), describing the cave before the major archaeological excavations, noted several examples of the Breccia having been higher, having pulled away from the overlying crust, leaving the roof overhead with bones sticking out of base of the calcite. The Water Gallery illustrates this. Figure 15 shows our reconstruction of the cross section at the Lake–Water Gallery, based on Pengelly's original field notes (1868–1880; still detailed at this early part of his excavations; however, they became less thorough as the dig progressed to other parts of the cave). Our date of 307 ± 5 ka (sample WGIU) places the lower layer of crystalline stalagmite in MIS 9. The dimensions and positions of the blocks of Crystalline Stalagmite distributed throughout the Breccia were carefully documented in Pengelly's notes (1868–1880). The thin horizontal slab of calcite just beneath the bone-rich layer of breccia may have been MIS 11, deposited in situ, and then fractured during the MIS 10 glacial period. The Breccia became mobilized during MIS 10, carrying MIS 11 bear bones from the Bear's Den toward the South Entrance, and was then covered by MIS 9 flowstone. (This sequence is corroborated by the evidence in the Bear's Den and Southwest Chamber, discussed next.) It is also interesting that the MIS 9 layer had not been fractured; the Breccia must have been intact when this thin layer was deposited. Some time after MIS 9 the Breccia was partly removed to create the Vacuity in the center of the passage. If the Breccia underneath it was removed before the next fracturing episode, then the MIS 9 layer would remain intact (see 4 of mechanisms of fracturing): this second episode of Breccia remobilization was thus probably in MIS 8. There is no evidence of fluvial activity in this movement of the Breccia: the redeposition of breccia plus bones farther down-passage still bears the marks of mass movement.

The Southwest Chamber is one of the best-documented passages because it was one of the first places to be excavated by Pengelly. We have reconstructed the cross section from Pengelly's notes (1868–1880), dated many of the calcite layers, and then reconstructed the most parsimonious suite of events required to explain the deposits (Fig. 16). The high-level remnants of MIS 11 and MIS 9 flowstone indicate that the Breccia must have been at least 2 m deep in order to provide a substrate on which the flowstone grew. Some of the Breccia must have been removed during MIS 10 in order to make space for the lower level MIS 9 calcite deposition. The Breccia must also have been quite mobile, because artifacts from the den area were transported up this passage embedded in the top of the Breccia. A mobile breccia is required until MIS 6a in order to incorporate the very securely dated SW1-O MIS 6b flowstone slab in the top of the Breccia. Another cracking event, during MIS 6a, was required to break the MIS 6b slab. Pengelly's notes (1868–1880) documented a thick layer of Crystalline Stalagmite across the passage; this must have represented the complex of deposition from MIS 11–5. The Breccia became indurated after MIS 6a, and further movement stopped.

Thus the evidence supports additional cracking events, during MIS 10 and MIS 6a, and remobilization of the Breccia during MIS 10, MIS 8, MIS 6c, and MIS 6a. Straw (1997; Figure 1 therein) mapped “detached Breccia pieces in Cave-Earth”; this implies that the Breccia was also cracked, probably during MIS 2 just before the Cave Earth moved from the Vestibule down toward the North and South Sally Ports. Some of the breccia in the Bear's Den appears to have been frozen, cracked, slightly faulted, and then cemented with calcite (not dated, but stratigraphically in line with the dated MIS 9 sample, suggesting another breccia-cracking event in MIS 10).

Mechanism for Fracture of the Flowstone Sheets

The fractured flowstone has triggered much speculation: MacEnery (1859) suggested tectonic activity as the cause, Pengelly (1876, p.176) suggested hydraulic pressure as the cause, and Proctor (1994) reverted to the earthquake theory. Straw (1997), based on Proctor's one date, assumed that it must have been caused by a single strong earthquake between ca. 100 ka and 75 ka. In response to Straw's (1997) article, Ford (1997) collected several references to fractures observed in other caves in Devon, and speculated that all of them are attributable to earthquake activity; however, he offered no substantive support for the fracture events. For example, Ford reassessed Sutcliffe's (1960) frost-heave interpretation of the fractured calcite sheet in Joint Mitnor cave as earthquake damage. The evidence Ford invoked is largely negative; he argued that frost heave is unlikely 100 km from the glacial margin and in a region under maritime influence, and, assuming that frost heave requires the presence of permafrost, the only explanation is earthquake damage.

We present unequivocal evidence that the fracturing occurred repeatedly throughout the middle to late Pleistocene history of the cave, and that it occurred during each cold episode since the first layer of calcite was deposited in MIS 11. The arguments that the region could not have been cold enough have very little empirical support. Croot and Griffiths (2001) mapped fossil polygons and stripes within ∼10 km of the cave and many examples of frost-related features both on Dartmoor and at many coastal sites all around Devon and Cornwall. Some reconstructions of conditions at 20 ka in the region (Murton and Lautridou, 2003) show the whole region, including Torquay and Dartmoor, under continuous permafrost, while others show permafrost only on Dartmoor (and presumably seasonal frost elsewhere), and some show discontinuous perma frost just a few kilometers north of Torquay. Evidence that Devon probably underwent discontinuous permafrost, at least during the coldest periods, was provided by Ballantyne and Harris's (1993) finding of a fossil pingo on West Dartmoor. Croot and Griffiths (2001) suggested that during glacial periods the climate of Devon may have been similar to that of Svalbard today (which displays excellent examples of peri-glacial activity and frost-related features).

Considering that frost heave does not require the presence of permafrost, either continuous or discontinuous, and that it only required one significant frost-heave event per cold interval to fracture the relatively thin calcite sheets, we argue that the evidence for frost shattering is overwhelming. The occurrence of at least seven earthquakes of sufficient magnitude to crack the sheets, all coincidental with glacial periods, is extremely improbable. As further argument we cite Forti's (2004) discussion of tectonic effects on speleothems: tectonic stresses can be recognized by the characteristic fracture, the most typical being breakage of stalagmites along sub-horizontal planes, by the consistent breakage in certain directions, and by breakages being grouped together in time. Forti (2004) was adamant that all other possible causes of breakage must be discounted, such as simple mechanical failure from increase in loading, from sliding of stalagmites and columns on unconsolidated materials, and from tongues of ice during glaciation. All Forti's tectonic examples are of broken stalagmites or columns; none are of flowstone sheets, and all examples come from regions of known and significant tectonic activity. We suggest that the subhorizontal shear operating on a flowstone sheet would produce low-angle greenstick fractures: the cracks we see in the cave are subvertical.

If frost heave is responsible for flowstone fracture, then there should be an association of the relative thickness of both the potentially heaving material and the potentially fracturing material. So, a thick layer of water-saturated breccia will expand more than a thin layer, and a thin layer of flowstone will fracture more easily than a thick layer. Figure 17A is a simple map of the Breccia thickness. Both Proctor (1994) and Straw (1997) mapped the distribution of the Breccia from Pengelly's original reports. This is the basis for Figure 17A, but we have divided the Breccia according to thickness, reconstructed partly from sections in Proctor (1994) and partly from field observations. Figure 17B shows the distribution of intact flowstone and fractured flowstone, a simplified version of Straw's (1997) map. The accuracy of this reconstruction is partly limited by the detail and accuracy of Pengelly's reports; we are aware that the level of detail deteriorated as his dig progressed, the earlier field notes being considerably more detailed than the later ones (1868–1880). For example, we are aware of some areas, such as the Labyrinth, that show evidence that Pengelly blasted through intact crystalline stalagmite, yet this is not shown in Straw's (1997) map. Figure 17C shows the distributions overlain: the two main areas of shattered flowstone (yellow) coincide largely with thick breccia (blue), producing the pattern of green in the Bear's Den and the Hedges Boss–Inscribed Boss region. The intact flowstone (red) in many areas is not underlain by any breccia: otherwise it is underlain by thin breccia (gray). The combination of thin breccia and fractured flowstone (gray-green) occurs only in a very thin line on the western side of Clinnick's Gallery. The combination of thick breccia and intact flowstone (purple) occurs only under the bosses, where the flowstone is too thick to fracture. Thus Figure 17C offers support for the assertion that the fracture is associated with the thickest breccia and is thus most likely to represent frost heave. Freezing of the Breccia could easily have been achieved in a single cold season, and sufficient aeration of the cave was likely, in view of the numerous open or semi-open routes for entry of sediments and/or animals.

Mechanism for Remobilization of Breccia

Evidence for removal and remobilization of the Breccia is clear. Some removal may have been from fluvial activity, but the only clear evidence of this is in the paragenetic activity on the roof of Rocky Chamber. The upper few decimeters of breccia has remained mobile such that artifacts and fractured pieces of calcite were moved from their source (e.g., Bear's Den) to new locations (e.g., Southwest Chamber) and yet are still emplaced within the Breccia matrix with no evidence for fluvial deposition. We reject the theory that water-rich debris flows effected this remobilization on the basis that little evidence can be found for fluvial orientation, sorting, or lamination. Long sections of the routes taken by the Breccia (Fig. 18) show slopes of ∼3°. Debris flows generally require slopes of >20°, but solifluction flows can operate on only a few degrees. The slopes of the entrance regions are steep enough to allow for either debris flow or solifluction. However, the gentle slopes of the main passages would preclude debris flows as a mechanism for breccia remobilization. We suggest that the most likely mechanism for Breccia and Cave Earth movement is solifluction.

The sedimentary history of Kents Cavern is considerably more complex than has previously been recognized. The deposits record a rich, cyclic history of events that track all the major climatic cycles of the past 500 k.y. of British Pleistocene history—a situation not known for any other European cave or subaerial site. Most of the deposits show some evidence of complex reworking, and the flowstone layer designated in the literature as Crystalline Stalagmite is shown to be a multilayered complex spanning MIS 11 to MIS 3. In summary, the first deposit is of fluvial sands. The second deposit, of muddy breccia, incorporates the famous Acheulian artifacts that were most likely fabricated during MIS 13 (latest Cromerian) and transported into the cave by natural processes through suffosion dolines during MIS12. Subsequently, each interglacial period produced calcite deposition in the cave, and each glacial period caused periglacial activity in the cave, during which the calcite was fractured by frost heave, incorporated into the mud, and the mud moved by solifluction.

The full sequence is as follows (see Fig. 19 and 03).

  1. MIS 15–13: late Cromerian warm stage: extensive surface weathering of hillside; deep regolith forms; vadose transport into cave of fluvial mud, sand and gravel, the Red Sands; fabrication of Acheulian artifacts.

  2. MIS 12: Anglian glaciation: formation and emplacement of breccia with entrained artifacts.

  3. MIS 12 (terminal phase)–early MIS 11: Bears hibernate in cave.

  4. MIS 11: Deposition of calcite layers; bears hibernate in cave.

  5. MIS 10: Cracking of flowstone and of breccia; removal and mobilization of breccia.

  6. MIS 9: Deposition of calcite.

  7. MIS 8: Cracking of calcite, remobilization of breccia.

  8. MIS 7: Deposition of calcite. 9. MIS 6c: Cracking of calcite.

  9. MIS 6b: Deposition of thin calcite flow-stone.

  10. MIS 6a: Cracking of flowstone, remobilization of breccia.

  11. MIS 5 e-a: Deposition of calcite. 13. MIS 4: Cracking of calcite; remobilization of breccia.

  12. MIS 3: Deposition of calcite, emplacement of Cave Earth.

  13. MIS 2: Cracking of calcite and breccia; mobilization of Cave Earth.

  14. MIS 1: Deposition of Granular Stalagmite.

In addition to providing the most complete Pleistocene record yet documented from any British cave, this report is also the first publication of well-dated and clearly documented evidence of frost heaving in interior cave passages. The magnitude of the internal cave response to major global climatic shifts is of interest. This history of repeated sedimentation events followed by frost shattering and remobilization events is probably unique in the karst literature. The uniqueness of the Kents Cavern sequence is likely an artifact of the relative lack of study that cave sedimentary sequences—rather than simply isolated speleothems—have received. We hope that the Kents cavern record will serve as a demonstration of the potential of these sequences, and focus future attention on the conservation and study of other potentially important cave sedimentary sequences in Britain and elsewhere.

We are indebted to Nick Powe, owner and managing director of Kents Cavern, for his encouragement of this project. We also thank Barry Chandler of the Torquay Museum for access to specimens and manuscripts in his care, and two anonymous referees for helpful comments on an earlier version of this manuscript. We are grateful to Darryl E. Granger and the staff of PRIME Lab, Purdue University, for cosmogenic dating support and advice. Permission to collect samples from this scheduled Ancient Monument was granted by English Heritage Secretary of State for Culture, Media and Sport under permit # HSD 9/2/7663. Partial funding was provided by the Andrew Mellon Foundation under a sabbatical improvement grant through Scripps College (to McFarlane), and through a research support grant from Pitzer College (to McFarlane), and Natural Sciences and Engineering Research Council grant (to Lundberg). This is Ottawa-Carleton Geo-science Centre, Isotope Geochemistry and Geochronology Research Facility contribution 44.